Predictive Modeling to Support Student Success on Step 2 Clinical Knowledge: A Multi-Cohort Model

Area of Science:

• Medical Education
• Assessment and Evaluation

Background:

• The United States Medical Licensing Examination (USMLE) Step 1 transition to pass/fail has increased focus on Step 2 Clinical Knowledge (CK) for residency admissions.
• Accurate early-warning systems are needed to identify medical students at risk of underperforming on the USMLE Step 2 CK.

Purpose of the Study:

• To develop and externally validate a concise predictive model for USMLE Step 2 CK performance.
• To guide targeted academic support for medical students midway through their clerkships.

Main Methods:

• A multiple linear regression (MLR) model was constructed using data from 422 students across three cohorts.
• National Board of Medical Examiners (NBME) subject exam scores were utilized as primary predictors.
• Predicted scores were converted to pass probabilities and stratified into low-risk and elevated-risk groups to trigger remediation.

Main Results:

• The MLR model explained 67% of the Step 2 CK score variance in-sample (R² = 0.673) and 61.2% out-of-sample.
• NBME subject exams in Medicine (r=0.703) and Pediatrics (r=0.693) were the strongest individual predictors.
• The risk stratification identified 9% of students for remediation; 99.1% of high-risk students who received support passed Step 2 CK.

Conclusions:

• A six-variable model using NBME shelf exam scores reliably forecasts USMLE Step 2 CK outcomes.
• This predictive tool provides actionable, early guidance for targeted interventions, conserving advising resources.
• Implementing this model within an equity-minded remediation framework significantly reduces the risk of Step 2 CK failure.